Comparative Analysis of Predictive Models for Crash Injury Severity Among Aging Drivers in Florida

Abstract: The increasing vulnerability of the aging population in traffic incidents highlights the need for effective predictive tools to assess crash injury severity. This research leveraged a dataset from Signal Four Analytics, documenting crash records from 2019 to 2023 involving drivers aged 65 and older on Florida’s freeways. The variables used were driver age, gender, vision obstruction, lighting condition, weather condition, freeway area location, posted speed limit, roadway alignment code, roadway grade code, vehicle age, vehicle maneuver, day of the crash, and crash time, which could influence crash outcomes.
The primary objective of this study was to evaluate the effectiveness of Ordinal Logistic Regression (OLR), K-Nearest Neighbors (KNN), and Random Forest (RF) in predicting injury severity in crashes. The results showed that OLR performed better than the machine learning models, achieving an Area Under the Receiver Operating Characteristic (AUROC) of 0.587, a prediction accuracy of 0.755, and an F1 score of 0.861. These findings emphasize OLR’s strength in managing the ordered categories of injury severity data for older drivers.
The analysis pinpointed several significant risk factors that correlated with increased crash injury severity. Drivers over the age of 75, those with vision obstructed, and crashes occurring in poor lighting conditions were identified as particularly susceptible to severe injuries. This insight draws attention to the serious risks older adults face under adverse driving conditions and the critical need for tailored safety measures.
Considering these findings, the study proposes several interventions to improve road safety for the aging demographic. These include offering specialized refresher driving courses for older adults, enhancing roadway lighting, and incorporating advanced vehicle safety technologies. Such initiatives are expected to address the primary risk factors identified, such as poor visibility and the challenges of navigating at high speeds.
The research suggests that future studies should investigate other factors that might affect crash severity for older drivers, like medication use and cognitive impairments. Including these aspects could help in gaining a clearer picture of the challenges aging drivers face, leading to better-informed safety strategies.
Ultimately, this study provides valuable insights into the dynamics of crash injury severity among older drivers and confirms the effectiveness of OLR in analyzing complex crash data. The findings are ready to inform policymakers and transportation planners in developing measures that enhance the safety and mobility of aging populations on roadways, contributing to safer driving environments for all.